Speech recognition systems may use voice activity detection to determine when to perform speech recognition. For example, the speech recognition system may detect voice activity in audio input and, in response, may determine to generate a transcription from the audio input.